Skip to Main content Skip to Navigation

Nouvelles méthodes pour l’apprentissage non-supervisé en grandes dimensions.

Abstract : Spurred by recent advances on the theoretical analysis of the performances of the data-driven machine learning algorithms, this thesis tackles the performance analysis and improvement of high dimensional data and graph clustering. Specifically, in the first bigger part of the thesis, using advanced tools from random matrix theory, the performance analysis of spectral methods on dense realistic graph models and on high dimensional kernel random matrices is performed through the study of the eigenvalues and eigenvectors of the similarity matrices characterizing those data. New improved methods are proposed and are shown to outperform state-of-the-art approaches. In a second part, a new algorithm is proposed for the detection of heterogeneous communities from multi-layer graphs using variational Bayes approaches to approximate the posterior distribution of the sought variables. The proposed methods are successfully applied to synthetic benchmarks as well as real-world datasets and are shown to outperform standard approaches to clustering in those specific contexts.
Document type :
Complete list of metadatas

Cited literature [135 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Tuesday, October 9, 2018 - 11:36:05 AM
Last modification on : Wednesday, September 16, 2020 - 4:48:48 PM
Long-term archiving on: : Thursday, January 10, 2019 - 2:05:46 PM


Version validated by the jury (STAR)


  • HAL Id : tel-01891093, version 1


Hafiz Tiomoko Ali. Nouvelles méthodes pour l’apprentissage non-supervisé en grandes dimensions.. Autre [cs.OH]. Université Paris-Saclay, 2018. Français. ⟨NNT : 2018SACLC074⟩. ⟨tel-01891093⟩



Record views


Files downloads